Sergey Nivens - Fotolia
As companies must deliver ever-higher levels of uptime, data center backup power supplies gain more attention. They have also grown in sophistication, and admins can use multiple system types and industry standards to ensure trouble-free operations.
According to the April Research and Markets report "Data Center Power Market - Global Outlook and Forecast 2020-2025," uninterruptible power supply (UPS) systems represent the largest revenue contributor to the growing power backup systems market. Market share is then split between generators, transfer switches and switchgears, and power distribution units.
Researchers noted that the wide adoption of modular, scalable and lithium-ion powered UPSes are a likely trend over coming years.
If continuous operations are the main goal, organizations can start with appropriately sizing the resources they will need, says Mark Acton, a U.K.-based data center consultant.
"You need to start by making sure the amount of backup capacity matches the amount of utility power coming in," he said. For example, if a data center takes 50 megawatts from a utility, teams should ideally have the same amount of backup power.
Organizations typically start out with a smaller amount of power than necessary and then increase it over time as experience dictates and budgets allow.
"Look at your utility bill and remember that if you try to draw more power than your backup systems are designed for, they will generally just shut down," Acton said. Furthermore, if infrastructure operates close to design limits, this usually shortens the power supply's lifespan.
Backup power standards for implementation and operation
Acton recommended that admins reference standards and industry planning documents to develop data center backup power infrastructure. These include the ANSI/TIA 942 standard, which specifies the minimum requirements for structured cabling work; and ANSI/BICSI 002-2019, which covers all aspects of data center design; as well as the European EN 50600 standard for assessing power needs for specific purposes.
The Uptime Institute also provides effective guidance to build out backup power systems with its Tiers of operational redundancy and support.
"Every data center must be self-sufficient for some time because utility power can always go away, but at Tier I, there is no further redundancy," said Christopher Brown, CTO at the Uptime Institute.
There are four Tiers for backup power support within the data center:
- Tier I is a basic data center with just enough features to support its IT load with any kind of on-site power.
- Tier II steps up capabilities and adds at least one more generator.
- Tier III provides for concurrent maintainability. All components are redundant and the power paths are also redundant.
- Tier IV is completely fault tolerant; the systems can support not only concurrent maintenance, but also autonomously responds to faults or failures at a component or distribution level.
Data center backup power system selection
Exactly what kind of systems are adopted to achieve those levels is another question entirely, Brown said. Typically, admins should start with a discussion of what infrastructure is critical enough to require backup power.
"The answer can potentially include life-safety functions such as lighting and HVAC, as well as direct support of IT equipment, and some of those decisions need to be based on local codes," he said.
Christopher BrownCTO, Uptime Institute
The first step is for admins to decide on a UPS type, as most infrastructure does not start up immediately without some sort of power -- with the exception of a fuel cell system.
But certain fuel cells can have problematic power quality, so it's not a guaranteed connected data center backup power component.
"[Fuel cells] may be good enough for mechanical systems like chillers, but for electronics, you want to have some filtering," Brown said. This makes the role of a UPS even more important.
To home in on an answer as to how much data center backup power an organization should have depends on specific infrastructure requirements. Different systems may require different power quality. Ride-through time -- how long an IT team can operate without utility power -- also varies across businesses.
The answer depends on factors such as data transfer to the cloud or another physical facility, in which case ride-through time must be sufficient enough to accomplish those transfers. Or an organization may need to operate indefinitely, as might be necessary in the case of a major natural disaster.
"At a bare minimum, any system needs enough UPS[es] to support an orderly shutdown that protects data and equipment," Brown said.
For interim loads, some organizations adopt flywheel-based energy storage systems, which can take on a load as needed. Brown warns that they typically only supply enough power for 14-18 seconds.
"Companies typically couple them with fast-start power systems that can be up and generating in 10 seconds or so," he said.
Go beyond the UPS
To ensure a level of speed and reliability, some organizations put redundant diesel generators in place with separate fuel supplies to reduce the need to have a generator that can come online quickly. Most systems also have oil and coolant maintained at close-to-operating temperature to increase the likelihood of a smooth start.
"The additional cost can be significant, but it must be weighed against the great cost of downtime," Brown said.
Other alternatives to diesel generators as data center backup power include internal combustion engines using natural gas or propane fuel, as well as gas turbines.
Brown said power backup options used to be very simple, but over the years, due to government regulations and environmental concerns, there is increased pressure to eliminate or reduce dependence on diesel generators; even batteries can come under scrutiny.
Despite the desire to get away from internal combustion power, this option still offers the most power and reliability in the smallest and most affordable package. Turbines are slower to start and don't respond as well to load changes.
"[Internal combustion engines] can be bad for the environment but, in practice, don't run very often. In the IT industry [there are] few companies that would build a data center without one," he added.
Then, no matter what system admins select, there is the care and maintenance, all of which are prone to mechanical or electrical failures. If a system is operational for an extended period, it must be maintained during operations -- and between uses. As part of regular operations, teams should test data center backup power systems and practice rapid shutdown protocols.